Ultrafest analysis
Stefano Coretta
10/08/2017
1 Import data
languages <- read_csv("voicing-effect/stimuli/languages.csv")## Parsed with column specification:
## cols(
## speaker = col_character(),
## language = col_character()
## )words <- read_csv("voicing-effect/stimuli/nonce.csv")## Parsed with column specification:
## cols(
## item = col_integer(),
## word = col_character(),
## ipa = col_character(),
## c1 = col_character(),
## c1phonation = col_character(),
## vowel = col_character(),
## anteropost = col_character(),
## height = col_character(),
## c2 = col_character(),
## c2phonation = col_character(),
## c2place = col_character(),
## language = col_character()
## )columns <- c(
"speaker",
"seconds",
"rec.date",
"prompt",
"label",
"TT.displacement.sm",
"TT.velocity",
"TT.velocity.abs",
"TD.displacement.sm",
"TD.velocity",
"TD.velocity.abs"
)
aaa.files <- list.files(
path = "./voicing-effect/results/ultrasound",
pattern = "*-tongue-cart.tsv",
full.names = TRUE
)
tongues <- read_aaa(
aaa.files,
columns,
na.rm = TRUE
) %>%
mutate(word = word(prompt, 2)) %>%
left_join(y = languages) %>%
left_join(y = words) %>%
mutate_if(is.character, as.factor) %>%
group_by(speaker) %>%
mutate(
X.re = rescale(X),
Y.re = rescale(Y)
) %>%
ungroup() %>%
mutate(
vowel.ord = ordered(vowel, levels = c("a", "o", "u")),
c2place.ord = ordered(c2place, levels = c("coronal", "velar")),
c2phonation.ord = ordered(c2phonation, levels = c("voiceless", "voiced"))
)## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character(),
## X_1 = col_character(),
## Y_1 = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Joining, by = "speaker"## Joining, by = c("word", "language")max <- tongues %>%
filter(label %in% c("max_TT", "max_TD"), vowel != "u") %>%
arrange(rec.date, fan.line) %>%
create_event_start("rec.date")
max_it_12 <- max %>%
filter(speaker %in% c("it01", "it02"))
max_pl_23 <- max %>%
filter(speaker %in% c("pl02", "pl03"))
max_pl_34 <- max %>%
filter(speaker %in% c("pl03", "pl04"))
aaa_files_clos <- list.files(
path = "./voicing-effect/results/ultrasound",
pattern = "*-tongue-clos-cart.tsv",
full.names = TRUE
)
tongues_clos <- read_aaa(
aaa_files_clos,
columns,
na.rm = TRUE
) %>%
mutate(word = word(prompt, 2)) %>%
left_join(y = languages) %>%
left_join(y = words) %>%
mutate_if(is.character, as.factor) %>%
group_by(speaker) %>%
mutate(
X.re = rescale(X),
Y.re = rescale(Y)
) %>%
ungroup() %>%
mutate(
vowel.ord = ordered(vowel, levels = c("a", "o", "u")),
c2place.ord = ordered(c2place, levels = c("coronal", "velar")),
c2phonation.ord = ordered(c2phonation, levels = c("voiceless", "voiced"))
) %>%
arrange(rec.date, fan.line) %>%
create_event_start("rec.date")## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character(),
## X_2 = col_character(),
## Y_2 = col_character(),
## X_3 = col_character(),
## Y_3 = col_character(),
## X_4 = col_character(),
## Y_4 = col_character(),
## X_5 = col_character(),
## Y_5 = col_character(),
## X_6 = col_character(),
## Y_6 = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character(),
## X_3 = col_character(),
## Y_3 = col_character(),
## X_4 = col_character(),
## Y_4 = col_character(),
## X_5 = col_character(),
## Y_5 = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Parsed with column specification:
## cols(
## .default = col_double(),
## speaker = col_character(),
## rec.date = col_character(),
## prompt = col_character(),
## label = col_character()
## )## See spec(...) for full column specifications.## Joining, by = "speaker"## Joining, by = c("word", "language")2 Exploration
max %>%
filter(c2place == "coronal", language == "italian") %>%
plot_tongue(geom = "point", alpha = 0.5) +
aes(colour = c2phonation) +
facet_grid(vowel ~ speaker)
max %>%
filter(c2place == "velar", language == "italian") %>%
plot_tongue(geom = "point", alpha = 0.5) +
aes(colour = c2phonation) +
facet_grid(vowel ~ speaker)
max %>%
filter(c2place == "coronal", language == "polish") %>%
ggplot(aes(X, Y, colour = c2phonation)) +
geom_point(alpha = 0.5) +
facet_grid(vowel ~ speaker) +
coord_fixed(1.5)
max %>%
filter(c2place == "coronal", language == "polish") %>%
ggplot(aes(X.re, Y.re, colour = c2phonation)) +
geom_point(alpha = 0.5) +
facet_grid(speaker ~ vowel) +
coord_fixed(0.7)
max %>%
filter(c2place == "velar", language == "polish") %>%
ggplot(aes(X, Y, colour = c2phonation)) +
geom_point(alpha = 0.5) +
facet_grid(vowel ~ speaker) +
coord_fixed(0.7)
max %>%
filter(c2place == "velar", language == "polish") %>%
ggplot(aes(X.re, Y.re, colour = c2phonation)) +
geom_point(alpha = 0.5) +
facet_grid(speaker ~ vowel) +
coord_fixed(0.7)
3 Testing
Comparing fixed effects models works best with ML, otherwise you can use fREML with AIC for fixed effects, but if there is an AR1 model cannot use AIC, so need ML.
3.1 Polish
pl.m1 <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_23,
method = "fREML"
)
rho <- start_value_rho(pl.m1)
pl.m1.ar <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_23,
method = "fREML",
rho = rho,
AR.start = max_pl_23$start.event
)
summary(pl.m1)
pl.m1.null <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_23,
method = "fREML"
)
compareML(pl.m1, pl.m1.null)plot_smooth(
pl.m1,
view = "X.re",
plot_all= "c2phonation.ord",
cond = list("c2place.ord" = "coronal"),
rug = FALSE
)
plot_diff(
pl.m1,
view = "X.re",
comp = list(c2phonation.ord = c("voiceless", "voiced")),
cond = list("c2place.ord" = "coronal")
)plot_gamsd(
pl.m1,
view = "X.re",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)plot_gamsd(
pl.m1.ar,
view = "X.re",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)3.1.1 PL02
pl02_max <- filter(max, speaker == "pl02", X > -20)
pl02.m1 <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl02_max,
method = "ML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(pl02.m1)
pl02.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl02_max,
method = "ML",
rho = rho,
AR.start = pl02_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(pl02.m1)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.23973 0.59721 5.425 6.45e-08 ***
## X -0.09449 0.03601 -2.624 0.00876 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.269 7.773 66.456 < 2e-16 ***
## s(X):c2phonation.ordvoiced 8.003 8.684 6.916 5.04e-07 ***
## s(X):c2place.ordvelar 8.877 8.983 309.316 < 2e-16 ***
## s(X):vowel.ordo 7.803 8.573 21.311 < 2e-16 ***
## s(X,rec.date) 188.143 300.000 9.608 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 342/343
## R-sq.(adj) = 0.949 Deviance explained = 95.4%
## -ML = 5542 Scale est. = 4.4805 n = 2391pl02.m1.null <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl02_max,
method = "ML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(pl02.m1, pl02.m1.null)## pl02.m1: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## pl02.m1.null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 pl02.m1.null 5557.001 11
## 2 pl02.m1 5541.991 13 15.011 2.000 3.027e-07 ***
##
## AIC difference: -35.85, model pl02.m1 has lower AIC.plot_gamsd(
pl02.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -19.967800 to 59.219800.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 07/02/2017 16:29:14.
3.1.2 PL03
pl03_max <- filter(max, speaker == "pl03")
pl03.m1 <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl03_max,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(pl03.m1)
pl03.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl03_max,
method = "ML",
rho = rho,
AR.start = pl03_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.pl03.m1.ar.null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl03_max,
method = "ML",
rho = rho,
AR.start = pl03_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(pl03.m1)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -5.15251 0.84915 -6.068 1.63e-09 ***
## X 0.42071 0.05257 8.004 2.35e-15 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.433 7.793 52.044 < 2e-16 ***
## s(X):c2phonation.ordvoiced 6.740 7.654 4.817 1.54e-05 ***
## s(X):c2place.ordvelar 8.755 8.938 33.136 < 2e-16 ***
## s(X):vowel.ordo 8.770 8.941 20.139 < 2e-16 ***
## s(X,rec.date) 209.072 235.000 26.599 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 277/278
## R-sq.(adj) = 0.969 Deviance explained = 97.3%
## fREML = 3995.5 Scale est. = 2.8743 n = 1793pl03.m1.null <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl03_max,
method = "ML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(pl03.m1.ar, pl03.m1.ar.null)## pl03.m1.ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## pl03.m1.ar.null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 pl03.m1.ar.null 3335.866 11
## 2 pl03.m1.ar 3330.848 13 5.018 2.000 0.007 **## Warning in compareML(pl03.m1.ar, pl03.m1.ar.null): AIC is not reliable,
## because an AR1 model is included (rho1 = 0.611727, rho2 = 0.611727).plot_gamsd(
pl03.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "velar")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -34.964200 to 70.078500.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): velar.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 24/03/2017 14:44:40.
3.1.3 PL04
filter(max, speaker == "pl04") %>%
ggplot(aes(X, Y, colour = c2phonation)) +
geom_point(alpha = 0.5) +
facet_grid(vowel ~ c2place)
filter(max, speaker == "pl04", X > -30) %>%
ggplot(aes(X, Y, colour = c2phonation)) +
geom_point(alpha = 0.5) +
facet_grid(vowel ~ c2place)
pl04_max <- filter(max, speaker == "pl04", X > -30)
pl04.m1 <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_max,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(pl04.m1)
pl04.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_max,
method = "ML",
rho = rho,
AR.start = pl04_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.pl04.m1.ar.null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_max,
method = "ML",
rho = rho,
AR.start = pl04_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(pl04.m1.ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.28296 0.38386 8.553 <2e-16 ***
## X -0.03567 0.03469 -1.028 0.304
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.538 7.847 31.856 < 2e-16 ***
## s(X):c2phonation.ordvoiced 1.006 1.009 0.296 0.589
## s(X):c2place.ordvelar 8.539 8.876 26.159 < 2e-16 ***
## s(X):vowel.ordo 7.975 8.620 10.694 7.37e-15 ***
## s(X,rec.date) 197.731 230.000 25.617 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 272/273
## R-sq.(adj) = 0.975 Deviance explained = 98%
## -ML = 1757.9 Scale est. = 0.72325 n = 1228pl04.m1.null <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_max,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.# compareML(pl04.m1, pl04.m1.null)
compareML(pl04.m1.ar, pl04.m1.ar.null)## pl04.m1.ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## pl04.m1.ar.null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 pl04.m1.ar.null 1758.008 11
## 2 pl04.m1.ar 1757.861 13 0.148 2.000 0.863## Warning in compareML(pl04.m1.ar, pl04.m1.ar.null): AIC is not reliable,
## because an AR1 model is included (rho1 = 0.518140, rho2 = 0.518140).## Warning in compareML(pl04.m1.ar, pl04.m1.ar.null): Only small difference in ML...plot_gamsd(
pl04.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -29.959700 to 32.556800.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 26/05/2017 17:58:11.
ggsave("presentations/2017-ultrafest/pl04.pdf", width = 6, height = 4)acf_plot(resid(pl04.m1.ar), split_by=list(pl04_max$rec.date))
acf_resid(pl04.m1.ar, split_pred = "AR.start")
pl04_a <- filter(max, speaker == "pl04", vowel == "a")
pl04.m1 <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_a,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(pl04.m1)
pl04.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_a,
method = "ML",
rho = rho,
AR.start = pl04_a$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.pl04.m1.ar.null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_a,
method = "ML",
rho = rho,
AR.start = pl04_a$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(pl04.m1.ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.83223 0.46029 1.808 0.0711 .
## X 0.02144 0.04680 0.458 0.6470
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.479 7.786 24.294 <2e-16 ***
## s(X):c2phonation.ordvoiced 3.384 4.190 0.553 0.575
## s(X):c2place.ordvelar 8.019 8.623 12.720 <2e-16 ***
## s(X,rec.date) 95.926 116.000 29.624 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 148/149
## R-sq.(adj) = 0.987 Deviance explained = 98.9%
## -ML = 1101 Scale est. = 0.95795 n = 689compareML(pl04.m1.ar, pl04.m1.ar.null)## pl04.m1.ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## pl04.m1.ar.null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 pl04.m1.ar.null 1101.707 9
## 2 pl04.m1.ar 1101.011 11 0.696 2.000 0.499## Warning in compareML(pl04.m1.ar, pl04.m1.ar.null): AIC is not reliable,
## because an AR1 model is included (rho1 = 0.455701, rho2 = 0.455701).## Warning in compareML(pl04.m1.ar, pl04.m1.ar.null): Only small difference in ML...If only /a/ is used, no significance at all. With /o/ and X > -30, no significance, but with all X some significant difference at the very left end of the tongue.
plot_gamsd(
pl04.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -41.892500 to 32.556800.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * rec.date : factor; set to the value(s): 26/05/2017 17:59:49.
3.1.4 PL03-PL04
pl34.m1 <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_34,
method = "fREML"
)
rho <- start_value_rho(pl34.m1)
pl34.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_34,
method = "ML",
rho = rho,
AR.start = max_pl_34$start.event
)
pl34.m1.ar.null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_34,
method = "ML",
rho = rho,
AR.start = max_pl_34$start.event
)
summary(pl34.m1.ar)
compareML(pl34.m1.ar, pl34.m1.ar.null)plot_gamsd(
pl34.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "velar", speaker = "pl04")
)The models with X, Y are not reliable because the tongue size differs a lot between speakers.
pl34.m1.re <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_34,
method = "fREML"
)
rho <- start_value_rho(pl34.m1.re)
pl34.m1.ar.re <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_34,
method = "ML",
rho = rho,
AR.start = max_pl_34$start.event
)
pl34.m1.ar.re.null <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_pl_34,
method = "ML",
rho = rho,
AR.start = max_pl_34$start.event
)
summary(pl34.m1.ar.re)
compareML(pl34.m1.ar.re, pl34.m1.ar.re.null)plot_gamsd(
pl34.m1.ar.re,
view = "X.re",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)
plot_gamsd(
pl34.m1.ar.re,
view = "X.re",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "velar")
)3.2 Italian
it01 and it02 have TRA at maximum displacement.
it.m1 <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_it_12,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(it.m1)
it.m1.ar <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_it_12,
method = "ML",
rho = rho,
AR.start = max_it_12$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(it.m1.ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y.re ~ X.re + s(X.re, bs = "cr") + s(X.re, by = c2phonation.ord,
## bs = "cr") + s(X.re, by = c2place.ord, bs = "cr") + s(X.re,
## by = vowel.ord, bs = "cr") + s(X.re, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5) + s(X.re, speaker, bs = "fs", xt = "cr",
## m = 1, k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.25090 0.06259 4.008 6.26e-05 ***
## X.re 0.80000 0.21532 3.715 0.000207 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X.re) 5.516 6.082 3.409 0.00394 **
## s(X.re):c2phonation.ordvoiced 5.106 6.209 5.491 1.33e-05 ***
## s(X.re):c2place.ordvelar 8.706 8.942 101.743 < 2e-16 ***
## s(X.re):vowel.ordo 5.286 6.405 8.554 1.45e-09 ***
## s(X.re,rec.date) 177.731 415.000 1.097 < 2e-16 ***
## s(X.re,speaker) 6.607 8.000 74.894 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 467/468
## R-sq.(adj) = 0.941 Deviance explained = 94.5%
## -ML = -6713.9 Scale est. = 0.0017165 n = 3165it.m1.ar.null <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
# s(X.re, by = c2phonation.ord, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_it_12,
method = "ML",
rho = rho,
AR.start = max_it_12$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.it.m1.null <- bam(
Y.re ~
X.re +
s(X.re, bs = "cr") +
s(X.re, by = c2place.ord, bs = "cr") +
s(X.re, by = vowel.ord, bs = "cr") +
s(X.re, rec.date, bs = "fs", xt = "cr", m = 1, k = 5) +
s(X.re, speaker, bs = "fs", xt = "cr", m = 1, k = 5),
data = max_it_12,
method = "ML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(it.m1.ar, it.m1.ar.null)## it.m1.ar: Y.re ~ X.re + s(X.re, bs = "cr") + s(X.re, by = c2phonation.ord,
## bs = "cr") + s(X.re, by = c2place.ord, bs = "cr") + s(X.re,
## by = vowel.ord, bs = "cr") + s(X.re, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5) + s(X.re, speaker, bs = "fs", xt = "cr",
## m = 1, k = 5)
##
## it.m1.ar.null: Y.re ~ X.re + s(X.re, bs = "cr") + s(X.re, by = c2place.ord,
## bs = "cr") + s(X.re, by = vowel.ord, bs = "cr") + s(X.re,
## rec.date, bs = "fs", xt = "cr", m = 1, k = 5) + s(X.re, speaker,
## bs = "fs", xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 it.m1.ar.null -6702.103 14
## 2 it.m1.ar -6713.863 16 11.761 2.000 7.805e-06 ***## Warning in compareML(it.m1.ar, it.m1.ar.null): AIC is not reliable, because
## an AR1 model is included (rho1 = 0.762798, rho2 = 0.762798).plot_gamsd(
it.m1.ar,
view = "X.re",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X.re : numeric predictor; with 100 values ranging from 0.000000 to 0.995850.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 29/11/2016 15:10:52.
## * speaker : factor; set to the value(s): it01.
3.2.1 IT01
it01_max <- filter(max, speaker == "it01")
it01.m1 <- bam(
Y ~
X.re +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it01_max,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(it01.m1)
it01.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it01_max,
method = "ML",
rho = rho,
AR.start = it01_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(it01.m1.ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -4.73141 0.30682 -15.42 <2e-16 ***
## X 0.76532 0.02666 28.70 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.764 7.944 181.554 < 2e-16 ***
## s(X):c2phonation.ordvoiced 5.037 6.136 9.480 2.06e-10 ***
## s(X):c2place.ordvelar 8.822 8.979 183.531 < 2e-16 ***
## s(X):vowel.ordo 6.862 7.808 11.566 1.19e-15 ***
## s(X,rec.date) 92.680 225.000 1.976 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 267/268
## R-sq.(adj) = 0.979 Deviance explained = 98%
## -ML = 3488.2 Scale est. = 4.0701 n = 1932it01.m1.ar.null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it01_max,
method = "ML",
rho = rho,
AR.start = it01_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(it01.m1.ar, it01.m1.ar.null)## it01.m1.ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## it01.m1.ar.null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 it01.m1.ar.null 3509.547 11
## 2 it01.m1.ar 3488.231 13 21.316 2.000 5.527e-10 ***## Warning in compareML(it01.m1.ar, it01.m1.ar.null): AIC is not reliable,
## because an AR1 model is included (rho1 = 0.745137, rho2 = 0.745137).plot_gamsd(
it01.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -40.731200 to 54.425500.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 29/11/2016 15:10:52.
ggsave("presentations/2017-ultrafest/it01.pdf", width = 6, height = 4)3.2.2 IT02
it02_max <- filter(max, speaker == "it02")
it02.m1 <- bam(
Y ~
X.re +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it02_max,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(it02.m1)
it02.m1.ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it02_max,
method = "ML",
rho = rho,
AR.start = it02_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(it02.m1.ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -1.67360 0.52437 -3.192 0.00145 **
## X 0.65159 0.03157 20.640 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 6.386 7.189 57.403 < 2e-16 ***
## s(X):c2phonation.ordvoiced 6.589 7.613 4.927 9.93e-06 ***
## s(X):c2place.ordvelar 8.630 8.910 52.382 < 2e-16 ***
## s(X):vowel.ordo 5.937 7.020 9.062 6.16e-11 ***
## s(X,rec.date) 72.997 185.000 1.401 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 227/228
## R-sq.(adj) = 0.916 Deviance explained = 92.2%
## -ML = 2803.6 Scale est. = 9.778 n = 1233it02.m1.ar.null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it02_max,
method = "ML",
rho = rho,
AR.start = it02_max$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(it02.m1.ar, it02.m1.ar.null)## it02.m1.ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## it02.m1.ar.null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 it02.m1.ar.null 2813.488 11
## 2 it02.m1.ar 2803.592 13 9.897 2.000 5.034e-05 ***## Warning in compareML(it02.m1.ar, it02.m1.ar.null): AIC is not reliable,
## because an AR1 model is included (rho1 = 0.735050, rho2 = 0.735050).plot_gamsd(
it02.m1.ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -44.949000 to 62.666500.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 12/12/2016 14:45:14.
4 Italian tongue at closure
tongues_clos %>%
filter(c2place == "coronal", language == "italian") %>%
plot_tongue(geom = "point", alpha = 0.5) +
aes(colour = c2phonation) +
facet_grid(vowel ~ speaker)
4.1 IT01
it01_clos <- tongues_clos %>%
filter(speaker == "it01", vowel != "u")it01_clos_m1 <- bam(
Y ~
X.re +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it01_clos,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(it01_clos_m1)
it01_clos_m1_ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it01_clos,
method = "ML",
rho = rho,
AR.start = it01_clos$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(it01_clos_m1_ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.61066 0.39198 1.558 0.119
## X 0.84888 0.02509 33.831 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.549 7.867 198.722 < 2e-16 ***
## s(X):c2phonation.ordvoiced 4.683 5.594 13.824 3.56e-14 ***
## s(X):c2place.ordvelar 8.641 8.924 174.974 < 2e-16 ***
## s(X):vowel.ordo 6.049 6.977 14.770 < 2e-16 ***
## s(X,rec.date) 96.251 230.000 1.789 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 272/273
## R-sq.(adj) = 0.981 Deviance explained = 98.2%
## -ML = 2983.9 Scale est. = 3.7241 n = 1648it01_clos_m1_ar_null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it01_clos,
method = "ML",
rho = rho,
AR.start = it01_clos$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(it01_clos_m1_ar, it01_clos_m1_ar_null)## it01_clos_m1_ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## it01_clos_m1_ar_null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 it01_clos_m1_ar_null 3014.072 11
## 2 it01_clos_m1_ar 2983.854 13 30.218 2.000 7.523e-14 ***## Warning in compareML(it01_clos_m1_ar, it01_clos_m1_ar_null): AIC is
## not reliable, because an AR1 model is included (rho1 = 0.721734, rho2 =
## 0.721734).plot_gamsd(
it01_clos_m1_ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -39.125800 to 54.502000.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 29/11/2016 15:10:52.
ggsave("presentations/2017-ultrafest/it01-clos.pdf", width = 6, height = 4)4.2 IT02
it02_clos <- tongues_clos %>%
filter(speaker == "it02", vowel != "u")it02_clos_m1 <- bam(
Y ~
X.re +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it02_clos,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(it02_clos_m1)
it02_clos_m1_ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it02_clos,
method = "ML",
rho = rho,
AR.start = it02_clos$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(it02_clos_m1_ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -2.56479 0.48605 -5.277 1.57e-07 ***
## X 0.64094 0.04079 15.714 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.047 7.612 40.215 < 2e-16 ***
## s(X):c2phonation.ordvoiced 7.191 8.131 5.061 2.41e-06 ***
## s(X):c2place.ordvelar 8.088 8.702 29.725 < 2e-16 ***
## s(X):vowel.ordo 8.257 8.776 10.910 2.49e-15 ***
## s(X,rec.date) 115.667 195.000 2.711 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 237/238
## R-sq.(adj) = 0.92 Deviance explained = 92.8%
## -ML = 2848.9 Scale est. = 6.7751 n = 1319it02_clos_m1_ar_null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = it02_clos,
method = "ML",
rho = rho,
AR.start = it02_clos$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(it02_clos_m1_ar, it02_clos_m1_ar_null)## it02_clos_m1_ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## it02_clos_m1_ar_null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 it02_clos_m1_ar_null 2861.022 11
## 2 it02_clos_m1_ar 2848.944 13 12.078 2.000 5.680e-06 ***## Warning in compareML(it02_clos_m1_ar, it02_clos_m1_ar_null): AIC is
## not reliable, because an AR1 model is included (rho1 = 0.718876, rho2 =
## 0.718876).plot_gamsd(
it02_clos_m1_ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -45.980000 to 64.280100.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): o.
## * rec.date : factor; set to the value(s): 12/12/2016 14:44:52.
Both it01 and it02 show TRA at closure. It is difficult to quantify the difference compared to TRA at maximum displacement. Need a model to check that.
5 Polish tongue at closure
5.1 PL04
pl04_clos <- tongues_clos %>%
filter(speaker == "pl04", vowel != "u")pl04_clos_m1 <- bam(
Y ~
X.re +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_clos,
method = "fREML"
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.rho <- start_value_rho(pl04_clos_m1)
pl04_clos_m1_ar <- bam(
Y ~
X +
s(X, bs = "cr") +
s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_clos,
method = "ML",
rho = rho,
AR.start = pl04_clos$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.summary(pl04_clos_m1_ar)##
## Family: gaussian
## Link function: identity
##
## Formula:
## Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -1.53763 0.11242 -13.677 <2e-16 ***
## X -0.01176 0.01746 -0.674 0.501
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(X) 7.817 7.951 234.521 < 2e-16 ***
## s(X):c2phonation.ordvoiced 1.004 1.005 0.809 0.369
## s(X):c2place.ordvelar 5.960 7.099 5.674 1.66e-06 ***
## s(X):vowel.ordo 7.493 8.406 16.701 < 2e-16 ***
## s(X,rec.date) 155.677 234.000 5.321 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Rank: 277/278
## R-sq.(adj) = 0.984 Deviance explained = 98.6%
## -ML = 1886.4 Scale est. = 0.83789 n = 1387pl04_clos_m1_ar_null <- bam(
Y ~
X +
s(X, bs = "cr") +
# s(X, by = c2phonation.ord, bs = "cr") +
s(X, by = c2place.ord, bs = "cr") +
s(X, by = vowel.ord, bs = "cr") +
s(X, rec.date, bs = "fs", xt = "cr", m = 1, k = 5),
data = pl04_clos,
method = "ML",
rho = rho,
AR.start = pl04_clos$start.event
)## Warning in gam.side(sm, X, tol = .Machine$double.eps^0.5): model has
## repeated 1-d smooths of same variable.compareML(pl04_clos_m1_ar, pl04_clos_m1_ar_null)## pl04_clos_m1_ar: Y ~ X + s(X, bs = "cr") + s(X, by = c2phonation.ord, bs = "cr") +
## s(X, by = c2place.ord, bs = "cr") + s(X, by = vowel.ord,
## bs = "cr") + s(X, rec.date, bs = "fs", xt = "cr", m = 1,
## k = 5)
##
## pl04_clos_m1_ar_null: Y ~ X + s(X, bs = "cr") + s(X, by = c2place.ord, bs = "cr") +
## s(X, by = vowel.ord, bs = "cr") + s(X, rec.date, bs = "fs",
## xt = "cr", m = 1, k = 5)
##
## Chi-square test of ML scores
## -----
## Model Score Edf Chisq Df p.value Sig.
## 1 pl04_clos_m1_ar_null 1886.770 11
## 2 pl04_clos_m1_ar 1886.368 13 0.401 2.000 0.669## Warning in compareML(pl04_clos_m1_ar, pl04_clos_m1_ar_null): AIC is
## not reliable, because an AR1 model is included (rho1 = 0.494196, rho2 =
## 0.494196).## Warning in compareML(pl04_clos_m1_ar, pl04_clos_m1_ar_null): Only small difference in ML...plot_gamsd(
pl04_clos_m1_ar,
view = "X",
comparison = list(c2phonation.ord = c("voiceless", "voiced")),
conditions = list(c2place.ord = "coronal")
)## Summary:
## * X : numeric predictor; with 100 values ranging from -39.613600 to 30.560600.
## * c2phonation.ord : factor; set to the value(s): voiced, voiceless.
## * c2place.ord : factor; set to the value(s): coronal.
## * vowel.ord : factor; set to the value(s): a.
## * rec.date : factor; set to the value(s): 26/05/2017 17:59:34.
ggsave("presentations/2017-ultrafest/pl04-clos.pdf", width = 6, height = 4)